Construction of Image Ontology using low-level features for Image Retrieval

This paper proposes a technique to construct Image Ontology using low-level features like color, texture and shape. The resulting ontology can be used to extract the relevant images from the image database. Retrieving relevant images from an image database is one of the challenging tasks in multimedia technology. More researches are being done in this area, among them Content-Based Image Retrieval (CBIR) is a note-worthy technique. Even in CBIR, the results retrieved are not satisfactory, i.e., with some test cases, there will be no relationship between queried image and retrieved images. It is difficult to produce a required result if the user query is not clear or when a user does not know what he wants. From a technical perspective, there are problems in producing a relevant result, if no image in the database has exact match with the user's requirement while using the low-level features. Interestingly, a semantic approach provides effective and meaningful results in image searching. Semantic searching approach uses Image Ontology for a better representation and organization of images. Besides the representation of the image properties, relationships among the several images can also be organized using image ontology. Low-level features reflect a direct visibility to the users. Therefore, construction of an Image Ontology using low-level features is more appropriate than the other techniques. In this paper an algorithm is proposed for constructing an image Ontology using low-level features.